A Q-learning based Electric Vehicle Scheduling Technique in a Distribution System for Power Loss Curtailment

Electric vehicles (EVs) are becoming more common in the distribution network (DN), which burdens the system throughout the charging cycle. However, if EVs are used to assist the utility according to its demands, the scenario turns profitable. In this paper, a new methodology to schedule the EVs and distributed generation (DG) according to the DN system demand is proposed. The proposed method is a hybrid of the intelligent Q-learning algorithm, Loss Sensitivity factor (LSF), and Enhanced Grasshopper Optimization (EGOA) techniques. The main objective of this work is to encourage peak shaving, minimize the power loss and improve the voltage profile of overall DN in presence of EVs. The modelling of EVs is done by considering the most significant parameters such as EV State of Charge (EV SoC), travel itinerary, capacity of EV battery, and charging/discharging levels. Initially the location of EV charging stations and DGs are determined using LSF technique. Further peak shaving of DN considering EVs behaviour is implemented using Q-learning based smart charging method. Later the sizes of DGs are determined using EGOA technique. EGOA is an enhancement in the already existing grasshopper optimization technique that improves the algorithm's exploring capability. The hybrid approach helped in identifying the strong buses for EV charging station and DG placement. Further a peak shave of overall DN demand is achieved using Q-learning technique. Later using EGOA the DG sizing along with optimal EV placement helped in reduction of power loss by around 67% compared to base case. The voltage profile of DN is enhanced with the proposed hybrid approach without using compensation devices. •A Q-learning-based EV scheduling algorithm is proposed by taking into account the most significant parameters such as EV State of Charge (EV SoC), travel itinerary, capacity of EV battery, and charging/discharging levels.•Strong nodes for placement of DGs and EVs are determined using LSF technique.•The authors propose an enhanced grasshopper algorithm (EGOA) for determining the size of DGs. This method is an advancement of the existing GOA algorithm.•The proposed method is used to minimize the power loss in a DN in the presence of uncertain EV clusters and DGs. This proposed approach combines Q-learning with LSF and EGOA techniques.